Which AHCI driver for SSD TRIM support?

I've been following the SSD mega thread for a couple weeks now.
Slowly the major SSD controllers have been getting TRIM support.
As of today Intel, Indilinx, and Samsung all support OS level TRIM functionality.

The problem is that drivers seem to be all over the place, which some groundless speculation as to whether or not certain drivers do pass the OS level TRIM commands to the drive.

Currently I have an i7 powered with laptop with the P55m chipset, and I have a Corsair P128 Samsung based SSD.

My driver options are currently:
-Default MS Drivers (apparently passes TRIM commands to drive)
-Intel Matrix Storage Drivers, does not pass TRIM to drives
-Intel Rapid Storage Drivers, these are apparently optimized for SSD drives but do not yet pass TRIM commands

Do I give up much performance with the MS driver, even though I do get TRIM? And how can you tell 100% for sure that TRIM is working. I keep reading about obscure bios settings blocking TRIM from working.

 

I would use the intel drivers... intel SSD's seem to own everything with these drivers so i would try them out...

 

Well, I wouldn't exactly call them 'obscure BIOS settings'.

You're talking about AHCI mode vs. IDE mode for SATA ports.

Your driver option list is correct afaik, but what I (think) I know is that the IRST drivers do pass trim to the drive too.

To test them is fairly straightforward I think. Create a partition on your fresh SSD and fill it to capacity with copied data - make sure you time it. Delete the data, repeat. Did they copy over at about the same speed/time (within 1 or 2% of each other)? Then TRIM is working. If, they take substantially longer, then TRIM is not doing it's job in your setup.

If I see a difference using the IRST drivers with a mechanical HD, I think you are giving up some performance on your SSD by using the default MS driver, but I would test this with real world work before committing myself that the IRST drivers are superior, by default.

Where do I see the 'better' with my 7K500 Hitachi and the IRST 9.5.4 drivers? Well, mostly the 'snap' the system has - but that too counts for something, right?

Cheers!

 

Performance... well, I believe some people had extremely high benchamrk results with Intel Rapid Storage Drivers - Intel Matrix vs. Microsoft vs. IDE Mode (MS Driver too I think) only seems to affect sequential speed for me - if you have to choose between these the MS driver is your best bet.

In the end its a "test it out" situation - someone force installed the Intel Rapid Storage Manager, got better results on an Intel SSD, I tried, the opposite happened, the Intel Matrix Drivers are quicker for me on Vista.

Or IDE mode if I wanted...

 

DetlevCM,

You do realize though that with the Intel Matrix Storage Manager drivers, you are not benefiting from TRIM support, right?

These drivers are known to not pass the TRIM commands to the SSD.

Cheers!

 

he doesn't have trim anyways on vista.

 

davepermen...

Duh! I guess I'm going off somewhere to hide now for a while...

 

Dave is right

And I think I told you before

I wouldn't gain anything from Windows 7 - in fact I would loose some nerves because Sony made a mess of some utilities... Fn keys to be exact.

 

I was thinking of the some power management Bios settings something like the acpi 2.0 settings.
I'm running windows 7, the intel rapid storage drivers seem to be working ok, but right now but my ATTO bench's have been going downhill week by week. At least the read speed stays high, and that's mostly why I got an SSD.

 

Then they might not support TRIM - download the Intel Toolbox if you have an Intel drive and manually trim it.

 

haha. don't hide, dude. just show public shame ^^

(and even on vista, he could use manual trim.. so while my statement was true, the conclusion of it wasn't, really )

 

Best way to know if TRIM is functional/enabled:

Command prompt > fsutil behavior query disabledeletenotify

DisableDeleteNotify = 1 (Windows TRIM commands are disabled)
DisableDeleteNotify = 0 (Windows TRIM commands are enabled)

Source: http://forums.legitreviews.com/about23670.html

 

this only makes sure it's enabled in windows. doesn't mean it's sure that the trim command gets propagated from the filesystem trough the sataport drivers trough the actual chip to the ssd, and get interpreted and handled there.

or does it?

 

I did this a while ago, and no it doesn't confirm the TRIM commands are actually making it to the drive. I suspect Intel is dragging their feet with a TRIM enabled AHCI drivers because they know their competition will benefit from it more than themselves. The Intel drives don't seem to show the same write speed degradation as the Samsung based drives.

 

superj,

You're right about that the Samsungs display the worst degradation - up to 70% according to Anand's third SSD anthology.

Have you tried the IRST drivers v9.5 to see if the TRIM is passed onto the SSD, at least an Intel G2, anyway?

Cheers!

 

I did run those drivers for a week or two recently here is the last benchmark I did with them installed. I'm back to the MS AHCI drivers now.

--------------------------------------------------
CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 203.871 MB/s
Sequential Write : 162.494 MB/s
Random Read 512KB : 158.664 MB/s
Random Write 512KB : 123.766 MB/s
Random Read 4KB : 11.935 MB/s
Random Write 4KB : 5.732 MB/s

Test Size : 100 MB
Date : 2010/01/12 8:25:04

 

Oh? Do the MS AHCI drivers perform better for you?

 

Here is a bench with the MS AHCI drivers. Note this was done later the same day as the first bench with the intel rapid storage drivers. I don't think I will see the full benefit of TRIM support until I write to almost the whole drive at least once to force the first round of garbage collection.

The 4k test showed the highest % change.

--------------------------------------------------
CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 215.586 MB/s
Sequential Write : 150.106 MB/s
Random Read 512KB : 167.786 MB/s
Random Write 512KB : 121.544 MB/s
Random Read 4KB : 15.913 MB/s
Random Write 4KB : 7.071 MB/s

Test Size : 100 MB
Date : 2010/01/12 18:10:27

 

Thanks superj.

I've always maintained that testing in our own systems should be 'mandatory' before we go blindly taking other's experience as mirroring our own.

Except for the Sequential Write, which is high enough anyways, it is a great improvement on all the other scores... nice!

Cheers!

 

i noticed there are big difference between winsat result and HD Tune Pro 8MB random seek (using version 4.01) for m$ driver vs IRSD 9.5 on intel G2. IRSD shows much better result than m$ driver although in AS SSD benchmark and CDM test m$ driver is a bit better.

 

vostro1400user,

yeah, benchmarks - ha!

I have to say, I tried the HD Tune Pro 4.01 and went back to 3.5.

4.01 seems very buggy to me (in my (clean Win 7 x64) install anyway) and it doesn't update the temps either (which is what I use it for mostly!).

I would be curious to see back to back runs of all the benchmarks you use posted. Just curious to see if there is a pattern or anything.

Thanking you in advance if you do decide to do this...

Cheers!

 

for winsat, the result shows below (m$ on left vs IRSD 9.5.0 on right), overall responsiveness mark for m$ is much worse.

for hdtune pro 4.01, things went weird. first time after switching back to m$ driver, i got 10 IOPS for 8M random seek while IRSD 9.5.0 shows 61 IOPS. i repeated three times for m$ driver tests i got same 10 IOPS. but now after i flash back to m$ driver again from IRSD 9.5.0, i got same 60-61 IOPS for 8M random seek, very strange.

for AS SSD benchmark, ms$ driver yields 370 while IRSD 9.5.0 got 348.

update: hdtune pro 8MB down to 29 now for m$ driver.

 

intel matrix and rapid storage drivers installed

CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 232.611 MB/s
Sequential Write : 85.417 MB/s
Random Read 512KB : 184.113 MB/s
Random Write 512KB : 84.340 MB/s
Random Read 4KB : 11.363 MB/s
Random Write 4KB : 27.270 MB/s

Test Size : 100 MB
Date : 2009/12/30 2:17:28




With only MS AHCI

--------------------------------------------------
CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 241.727 MB/s
Sequential Write : 85.668 MB/s
Random Read 512KB : 176.305 MB/s
Random Write 512KB : 84.996 MB/s
Random Read 4KB : 18.223 MB/s
Random Write 4KB : 43.305 MB/s

Test Size : 100 MB
Date : 2010/01/14 1:45:26

 

@dazz87:try to compare winsat results to see if there is any big difference upon overall responsiveness.

 

will try winsat later on tonight. For some reason, this new ssd does not feel snappy then my previous ssd. I have less program installed on this new drive, and the bootup time is about 8 seconds longer then the previous drive. With the intel toolbox when i first ran it with my old drive it took about 10 mins to finish. With this drive it took over 30 mins. First drive had both matrix and rapid storage drives installed where the new one does not. On the old drive i had windows 7 ultimate (build 7100) on the new drive I had Windows 7 Pro full.

 

Hi all

Can I just confirm something, in that the OS doesn't dictate TRIM support per se? It is the storage drivers alone, right?

I have an X25 G2, using Vista in AHCI mode, and I'm currently using intel's official RST driver v9.5.0.1037. In time, When Intel actually release RST drivers with TRIM support (going by one of the intel reps/users on their forum who indicated they will, in time), I'll be able to upgrade and benefit from (automatic) TRIM support, right, even though I'm using Vista?

Thanks.

 

At the moment only the MS AHCI drivers on Windows 7 allow for trimming while you use it.

With any other driver or any other OS you have to use the Intel toolbox to manually trim - but don't worry, it'll take a few months until you notice any slowdowns.

 

Thanks DetlevCM.

But basically, there's nothing stopping say, Intel from coming out and releasing their own storage drivers with TRIM support, and able to be used on an older OS like Vista?

 

I have a feeling that it would require some changes to the core of the OS.
Davepermen knows much more about it.

But as far as I am aware there will be no automatic trim for older OSes.

However - as I said above, it will take a long time before you notice it, and you can get the toolbox and schedule a manual trim

 

ahh. I think i understand now. I read some more and it does appear to rely on file system (NTFS) supporting it - something which Win 7 does, and vista doesn't.

Oh well, no biggy. as you said manually trimming (in this case, the intel tool schedules it) is pretty much like the old way... you'd have to defrag your old spindle drive... now it's you have to optimise your SSD

 

Anyone know which driver manufactures (e.g. Lenovo, Dell) use? They do offer SSD in their configured models, right?

 

Just to state it for the record. Intel has released Rapid Storage Drivers that support TRIM on all TRIM enabled SSDs (not just Intel). Not sure about RAID setups yet.

http://www.intel.com/support/chipsets/imsm/